Repository navigation
[css-syntax-3] The + in <an+b> confuses parsing tools #9473
Description
The <an+b> type is defined with a production rule that starts with <an+b> = .... This type is now referenced from the definition of :nth() in the CSS Generated Content for Paged Media Module.
Problem is that the + confuses grammar parsers, including CSSTree and the highlighting mechanism in Bikeshed (which creates a link to the definition of +), as they parse + as a component value multiplier.
The Value Syntax Definition is not entirely clear about what characters can be used to define basic data types between < and > but it says:
Other non-keyword literal characters that may appear in a component value, such as “+”, must be written enclosed in single quotes.
I think that's the only occurrence in CSS specs of a construct that parsers find problematic. Tools could be fixed of course, I'm wondering whether it would be ok to defined the type as <an-plus-b> instead.
(If so, I'm happy to prepare a PR to update the specs that define or reference the type)
Activity
cdoublev commentedon Oct 16, 2023
I think those tools should be able to parse any
<ident>between<and>, like in thesyntaxdescriptor.https://drafts.css-houdini.org/css-properties-values-api-1/#consume-a-data-type-name
tabatkins commentedon Oct 17, 2023
I think it's fine to just rename it to
<an-b>, honestly. The plus isn't necessary.[css-syntax-3] Rename `<an+b>` to `<an-b>`
fantasai commentedon Nov 20, 2023
Question about bikeshedding: as we're getting further away from the actual form of the notation "2n+1", it's getting more confusing to refer to the notation. An+B was somewhat understandable, because it pattern-matches against the notation. But
<an-b>just seems like weird 1970s computer garbage naming.Some possible replacements:
<An-B>at least helps it parse as three components (a, n, b), rather than two (an, b).<nth>or<ith>or<index>or something like that (it is called:nth-child())cdoublev commentedon Nov 21, 2023
<position-pattern>.I assume escaping
+inan+b, which would allow to be a valid<ident>parsed by CSSTree, is excluded.tabatkins commentedon Nov 21, 2023
Yeah, we need to name that's useful to read and write; there's no reason to contort ourselves to preserve a specific name. ^_^
Psychpsyo commentedon Mar 15, 2026
I like
<a-n-plus-b>and<nth>the most. Maybe<nth-notation>so that it at least feels like a noun.Anything that involves fully lowercase
anreads like the word "an" which feels weird and a bit ugly to read to me.<nth>or<nth-notation>would also acknowledge the fact that these can be of the form An-B as well. (or any other form that isn't strictly An+B)css-meeting-bot commentedon Jun 10, 2026
The CSS Working Group just discussed
[css-syntax-3] The `+` in `<an+b>` confuses parsing tools, and agreed to the following:RESOLVED: Use <a-n-plus-b> instead of <an+b>The full IRC log of that discussion
<dshin-moz> TabAtkins: Quick one, to check if everyone is ok with name<dshin-moz> ... Some selectors use an+b syntax, production for that is <an+b>
<dshin-moz> ... A lot of tools downstream of bikeshed do not like + in it
<dshin-moz> ... They want some sort of identifier
<dshin-moz> ... alphas, numerics & dashes
<TabAtkins> <a-n-plus-b>
<dshin-moz> ... I suggest we just write out the plus
<emilio> +1
<dshin-moz> ... Don't want anything abstract
<smfr> did you mean `<an-plus-b>` ?
<dshin-moz> ... Unless anyone have objections, I'll make the change to use it
<dbaron> I think I prefer Tab's <a-n-plus-b>
<dshin-moz> ... Prefer a-n to an, since that is a valid word
<dshin-moz> ... No strong opinions, though
<smfr> no opinion myself
<dbaron> (I'd also be ok with <anplusb>.)
<dshin-moz> PROPOSED: Use <a-n-plus-b> instead of <an+b>
<dshin-moz> Rossen: objections?
<dshin-moz> RESOLVED: Use <a-n-plus-b> instead of <an+b>
<schenney> present-
[various] Rename <an+b> to <a-n-plus-b> #9473
fantasai commentedon Jul 22, 2026
Re-opening because I think
<a-n-plus-b>is just the most awkward option out of all of the ones suggested in this issue, and I'd like us to consider one of the less awkward ones. 😓tabatkins commentedon Jul 22, 2026
While I don't have a strong opinion on what we use, I don't think a-n-plus-b is particularly bad - it's exactly how we pronounce the construct already (like, we say "nth-child starts with an 'A N plus B' argument"). It's also not too long; if anything, it's on the shorter side character-wise. I think it's just striking you because, tho short, it's still four words, and one of them is us spelling out a character.
I think we should stick with the resolution. It's how we all pronounce
An+Bwhen actually talking about it, and it's always convenient when the production is the same term as the actual definition.kbabbitt commentedon Jul 22, 2026
+1 that I don't think
<a-n-plus-b>is particularly bad. I think<nth>or<nth-index>would also be fine, but I'd prefer to avoid<An-B>or other forms that rely on casing to convey meaning.svgeesus commentedon Jul 30, 2026
I agree, it seems like a fine name to me for that reason.
tabatkins commentedon Oct 12, 2026
I'm gonna go ahead and close this; I've pinged Elika privately without response, but she can open a fresh issue if she really feels strongly enough to put it in front of the WG again.